From autonomous-dev-flow
Execute a single phase from roadmap to working code autonomously
How this skill is triggered — by the user, by Claude, or both
Slash command
/autonomous-dev-flow:skills/autonomous-phase-executionThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Execute one roadmap phase: think → plan → implement → handoff.
Execute one roadmap phase: think → plan → implement → handoff.
Receives from main executor:
Evaluate 2-3 architecture approaches internally:
Use decision framework:
Break phase into TDD task pairs:
For each task in sequence:
Write test
Verify RED
Write implementation
Verify GREEN
Run quality gates
Commit
Stage test and impl files
Commit with descriptive message:
feat: [component/feature name]
- What: [what was built]
- How: [pattern/approach used]
- Integration: [dependencies]
- Testing: [what tests cover]
After all tasks:
Create handoff document at docs/handoffs/phase-[N]-handoff.yml:
built: [Component1, Component2]
api:
- Component1.method(args)->return
- Component2.method(args)->return
patterns: [MVC, Repository]
Include:
built: List of components/modules created (names only)api: Key public interfaces (function signatures)patterns: Architectural patterns usedKeep handoff minimal: 50-100 tokens.
Before any commit, ALL must pass:
If any gate fails:
Python:
pytest tests/ -vruff check . && mypy .Go:
go test ./... -race -covergolangci-lint run ./...go build ./...Rust:
cargo testcargo clippy --all-targetscargo buildTypeScript:
npm testeslint . && tsc --noEmittscJava:
mvn test or gradle testmvn checkstyle:checkmvn compileWhen choosing architecture/technology:
Avoid:
Prefer:
If quality gate fails after 3 fix attempts:
If success criteria not met:
After completion, report:
npx claudepluginhub yarlson/claude-plugins --plugin autonomous-dev-flowExecutes implementation plans phase-by-phase: dispatches subagents per task, reviews once per phase with code-review skill, loads phases just-in-time, prints full outputs for transparency.
Enforces a gated Spec → Plan → Build → Test → Review → Ship lifecycle for multi-file features and projects, preventing AI coding agents from skipping specification and verification steps.
Executes scope-aware implementation with TDD and quality gates, scaling from inline plans for small changes to full multi-task workflows for larger work.